U.S. patent number 7,355,965 [Application Number 11/125,683] was granted by the patent office on 2008-04-08 for apparatus and method for rapid detection of unidirectional breaks in a network ring.
This patent grant is currently assigned to Overture Networks, Inc.. Invention is credited to David Stephen Griswold, Robert Leroy Lynch, Prayson Will Pate, Glenn Russell Swope.
United States Patent |
7,355,965 |
Griswold , et al. |
April 8, 2008 |
**Please see images for:
( Certificate of Correction ) ** |
Apparatus and method for rapid detection of unidirectional breaks
in a network ring
Abstract
Normal 802.3 Ethernet requires a tree topology. If a ring or a
loop exists, then packets will be forwarded around the ring
indefinitely. If the ring is broken, then there is no possibility
of packets being propagated forever. This invention shows how to
quickly impose a virtual break in the ring such that all nodes can
communicate with each other, and how to remove the virtual break
when a real failure occurs. This is accomplished by placing
intelligent nodes on the ring that work together to virtually break
and restore the ring. An embodiment is disclosed that handles a
unidirectional break in a communication link where the
unidirectional break is not sensed as an OPER DOWN state. This
abstract is provided as an aid to those performing prior art
searches and not a limitation on the scope of the claims.
Inventors: |
Griswold; David Stephen (Chapel
Hill, NC), Pate; Prayson Will (Durham, NC), Lynch; Robert
Leroy (Raleigh, NC), Swope; Glenn Russell (Raleigh,
NC) |
Assignee: |
Overture Networks, Inc.
(Morrisville, NC)
|
Family
ID: |
36648025 |
Appl.
No.: |
11/125,683 |
Filed: |
May 10, 2005 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20050201409 A1 |
Sep 15, 2005 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
60640278 |
Dec 31, 2004 |
|
|
|
|
60490764 |
Jul 29, 2003 |
|
|
|
|
60468325 |
May 6, 2003 |
|
|
|
|
Current U.S.
Class: |
370/216;
370/403 |
Current CPC
Class: |
H04L
12/42 (20130101); H04L 12/437 (20130101) |
Current International
Class: |
G01R
31/08 (20060101) |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
0052390 |
|
Oct 1981 |
|
EP |
|
0494695 |
|
Jul 1992 |
|
EP |
|
0519712 |
|
Dec 1992 |
|
EP |
|
0403763 |
|
Sep 1994 |
|
EP |
|
0472380 |
|
Jun 1999 |
|
EP |
|
07095227 |
|
Apr 1995 |
|
JP |
|
Other References
IEEE Standard 802.1D 1998 Edition, "Information
Technology--Telecommunications and Information Exchange Between
Systems--Local and metropolitan Area Networks--Common
Specifications--Part 3: Media Access Control (MAC) Bridges".
ISO/IEC 15802-3; ANSI/IEEE Std 802.1D, Sponsored by the LAN/MAN
Standards Committee of the IEEE Computer Society, pp. cover to xix,
50-54, and 78-109. cited by other .
Network Working Group Internet Draft Eric C. Rosen et al.,
"Multiprotocol Label Switching Architecture:
draft-ietf-mpls-arch-05.txt" (Apr. 1999) Title Page, Table of
Contents, .sctn. 1.1, .sctn. 2.23, and .sctn. 2.24. cited by other
.
RFC 3034 A. Conta et al., "Use of Label Switching on Fram Relay
Networks--Specification" (Jan. 2001), Copyright The Internet
Society (2001), Title Page, Introduction .sctn. 5.4, .sctn. 5.4.1,
and .sctn. 5.4.2; http://www.apps.ietf.org/rfc/rfc3034.html. cited
by other .
IEEE Standard 802.1w-2001 "IEEE Standard for Local and metropolitan
area networks--Common Specifications--Part 3: Media Access Control
(MAC) Bridges--Amendment 2: Rapid Reconfigurtion", Sponsored by the
LAN/MAN Standards Committee of the IEEE Computer Society. cited by
other .
Web pp. 1-10 printed May 10, 2004 entitled "How Ring Redundancy
Works" (2001):
http://industrialnetworking.com/support/tutorials/How.sub.--Ring.-
sub.--Redundancy.sub.--Works.html. cited by other .
Protected Switching Ring (PSR), Overture Networks, Inc. Feb. 6,
2004 (5 pages); OvertureNetworks.com/products/tech.sub.--psr.shtml.
cited by other .
Empower leagacy SONET/SDH and TDM networks for metro Ethernet
Revenues, Overture Networks, Mar. 14, 2003;
http://www.overturenetworks.com. cited by other .
Overture Expands Access Options for Smaller Firms, Jim Duffy,
Network World, Jan. 26, 2004;
www.nwfusion.com/News/2004/0126overture.html. cited by other .
Press Release--"Allied Telesyn Ships First True IP-based Platform
for Native POTS, FTTx, and ADSL2+" (Jun. 22, 2004), available over
Internet at:
http://www.alliedtelesyn.com.hk/press/news.sub.--20040622b.sub.--us.h-
tm. cited by other .
Allied Telesyn Technical Brief, "Active vs. PON: FTTx Technology
Choices" (Jun. 30, 2004), Rev. A, 10 pages, available at:
http://www.alliedtelesyn.com/corporate/media/whitepapers/active.sub.--vs.-
sub.--pon.sub.--a.sub.--wp.pdf. cited by other .
Allied Telesyn White Paper, "Ethernet Protection Switched Rings:
Creating Survivable Ethernet Network", Rev. B (Jul. 9, 2004), 12
pages, available on Internet at: http://www.allied
telesyn.com/corporate/media/whitepapers/epsr.sub.--wp.sub.--b.pdf-excerpt-
. cited by other .
Extreme Networks' Ethernet Automatic Protection Switching (EAPS)
Versio 1, S. Shah & M. Yip, publishe3d as RFC 3619 by Network
Working Group, www.ietf.org/rfc/rfc3619.txt, Oct. 2003. cited by
other .
Fast Reroute Extensions to RSVP-TE for LSP Tunnels
(draft-ietf-mpls-rsvp-1sp-fastreroute-01.txt); P. Pan et al.;
Expiration Dec. 2003, published as Internet-Draft by Internet
Engineering Task Force (IETF). cited by other .
Fast Reroute Extensions to Constraint Based Routed Label
Distribution Protocol (draft-vijay-mpls-crldp-fastreroute-02.txt);
Vijayanand.; Expiration May 2003, published as Internet-Draft by
Internet Engineering Tast Force (IETF). cited by other .
"Industrial Fiber Optic Networking", Publication: RM0900515 Rev.
Apr. 2004; Weed Instrument Company, Inc., (36 pages)
www.weedinstrument.com Copy was provided in priority U.S. Appl. No.
10/839,8964 (MPEP 609(d)). cited by other.
|
Primary Examiner: Wilson; Robert W.
Attorney, Agent or Firm: The Eclipse Group LLP Flynn; Kevin
E.
Parent Case Text
This application incorporates by reference and claims priority to
U.S. patent application No. 10/839,864, subsequently issued as U.S.
Pat. No. 6,928,059. This application also claims priority to two
applications claimed as priority documents in the '864 application
(U.S. Provisional Application 60/490,764 filed Jul. 29, 2003 and
U.S. Provisional Application 60/468,325 filed May 6, 2003). This
application incorporates by reference those two provisional
applications. Finally, this application claims priority to
co-pending U.S. provisional, application 60/640,278 filed Dec. 31,
2004 for Multipoint Protected Switching Ring and also incorporates
that application by reference.
Claims
What is claimed is:
1. A ring relay for use in a bidirectional network ring for a
bidirectional network ring that is manipulated to provide a virtual
break in order to operate within a network requiring a tree
topology, the bidirectional network ring carrying data packets; the
ring relay comprising the following attributes: having a first ring
port in data communication with a first incoming link and a first
outgoing link and a second ring port in data communication with a
second incoming link and a second outgoing link connected to the
bidirectional network ring; operating to pass HELLO messages
received on the first incoming link to the second outgoing link and
to pass HELLO messages received on the second incoming link to the
first outgoing link; and after detecting an OPER DOWN failure on
the second incoming link, reacting to the detection of the OPER
DOWN failure on the second incoming link by: imposing a break on
the second incoming link so that user data packets are blocked from
passing into the ring relay but the imposed break does not block
any incoming HELLO messages; imposing a break on the second
outgoing link so that user data packets are blocked from leaving
the ring relay on that link; sending a type of HELLO message and a
control signal out the second outgoing link such that an adjacent
node receives a HELLO message to provide assurance that the second
outgoing link from that ring relay is operating and to convey to a
node other than the ring relay that the ring relay is imposing a
bidirectional break so that the virtual break should be removed;
detecting the cessation of the OPER DOWN failure on the second
incoming link then resuming operating to pass HELLO messages
received on the first incoming link to the second outgoing link;
and delaying the removal of the imposed break for user data packets
on the second outgoing link and on the second incoming link for a
time period sufficient to allow the bidirectional network ring to
impose a virtual break at a node other than the ring relay.
2. The ring relay of claim 1 wherein the action of sending a type
of HELLO message and a control signal out the second outgoing link
such that an adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is accompanied by sending a type of
HELLO message and a control signal out the first outgoing link such
that an adjacent node receives a HELLO message to provide assurance
that the first outgoing link from that ring relay is operating and
to convey to a node other than the ring relay that the ring relay
is imposing a bidirectional break so that the virtual break should
be removed.
3. The ring relay of claim 1 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is achieved by sending a type of
HELLO message that includes an indication that the ring relay is
imposing a bidirectional break.
4. The ring relay of claim 1 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is achieved by sending a series of
control messages not part of a HELLO message wherein the series of
control messages convey the control signal to the node other than
the ring relay to remove the virtual break and keep the virtual
break removed as long as the series of control messages are
received by the node other than the ring relay.
5. The ring relay of claim 1 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is achieved by sending a control
message not part of a HELLO message wherein the control message
conveys the control signal to the node other than the ring relay to
remove the virtual break until another control message is sent
conveying the need to re-impose the virtual break.
6. A ring relay for use in a bidirectional network ring for a
bidirectional network ring that is manipulated to provide a virtual
break in order to operate within a network requiring a tree
topology, the bidirectional network ring carrying data packets; the
ring relay comprising the following attributes: having a first ring
port in data communication with a first incoming link and a first
outgoing link and a second ring port in data communication with a
second incoming link and a second outgoing link connected to the
bidirectional network ring; operating to pass HELLO messages
received on the first incoming link to the second outgoing link and
to pass HELLO messages received on the second incoming link to the
first outgoing link; and after detecting a gap in the receipt of
HELLO messages on the second incoming link in excess of a set
threshold, reacting by: imposing a break on the second incoming
link so that user data packets are blocked from passing into the
ring relay but the imposed break does not block any incoming HELLO
messages; imposing a break on the second outgoing link so that user
data packets are blocked from leaving the ring relay on that link;
sending a type of HELLO message and a control signal out the second
outgoing link such that an adjacent node receives a HELLO message
to provide assurance that the second outgoing link from that ring
relay is operating and to convey to a node other than the ring
relay that the ring relay is imposing a bidirectional break so that
the virtual break should be removed; after receiving a HELLO
message on the second incoming link, resuming operating to pass
HELLO messages received on the first incoming link to the second
outgoing link; and delaying the removal of the imposed break for
user data packets on the second outgoing link and on the second
incoming link for a time period sufficient to allow the
bidirectional network ring to impose a virtual break at a node
other than the ring relay.
7. The ring relay of claim 6 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that an adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is accompanied by sending a type of
HELLO message and a control signal out the first outgoing link such
that an adjacent node receives a HELLO message to provide assurance
that the first outgoing link from that ring relay is operating and
to convey to a node other than the ring relay that the ring relay
is imposing a bidirectional break so that the virtual break should
be removed.
8. The ring relay of claim 6 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is achieved by sending a type of
HELLO message that includes an indication that the ring relay is
imposing a bidirectional break.
9. The ring relay of claim 6 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is achieved by sending a series of
control messages not part of a HELLO message wherein the series of
control messages convey the control signal to the node other than
the ring relay to remove the virtual break and keep the virtual
break removed as long as the series of control messages are
received by the node other than the ring relay.
10. The ring relay of claim 6 wherein the act of sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring relay is
operating and to convey to a node other than the ring relay that
the ring relay is imposing a bidirectional break so that the
virtual break should be removed is achieved by sending a control
message not part of a HELLO message wherein the control message
conveys the control signal to the node other than the ring relay to
remove the virtual break until another control message is sent
conveying the need to re-impose the virtual break.
11. The ring relay of claim 6 wherein the delaying of the removal
of the imposed break for user data packets on the second outgoing
link and on the second incoming link for a time period sufficient
to allow the bidirectional network ring to impose a virtual break
at the node other than the ring relay is accomplished by delaying
the removal of the imposed break for user data packets until a
control signal is received from the node other than the ring relay
that the virtual break has been imposed.
12. A ring device for use in a bidirectional network ring for a
bidirectional network ring that is manipulated to provide a virtual
break in order to operate within a network requiring a tree
topology, the bidirectional network ring carrying data packets; the
ring device: having a first ring port in data communication with a
first incoming link and a first outgoing link and a second ring
port in data communication with a second incoming link and a second
outgoing link connected to the bidirectional network ring; and
detecting an absence of HELLO protocol packets on the second
incoming link in excess of a specified time duration then: blocking
the second outgoing link for user data packets such that a virtual
break is imposed on the second outgoing link; generating and
sending a series of modified HELLO protocol packets on the second
outgoing link, the sending of modified HELLO protocol packets onto
the second outgoing link used to trigger immediate corrective
action by the bidirectional network ring to remove an existing
virtual break from a device other than this ring device; and
detecting a HELLO protocol packet on the second incoming link then:
ceasing the generation and sending of the series of modified HELLO
protocol packets on the second outgoing link; and delaying the
removal of the blocking of user data packets on the second outgoing
link for a time period sufficient to allow the bidirectional
network ring to impose a virtual break at a node other than this
ring device.
13. The ring device of claim 12 wherein the act of generating and
sending a series of modified HELLO protocol packets on the second
outgoing link is accompanied by generating and sending a series of
modified HELLO protocol packets on the first outgoing link and the
act of ceasing the generation and sending of the series of modified
HELLO protocol packets on the second outgoing link is accompanied
by ceasing the generation and sending of the series of modified
HELLO protocol packets on the first outgoing link.
14. A slave arbiter for use in a bidirectional network ring for a
ring that is manipulated to provide a virtual break in order to
operate within a network requiring a tree topology, the
bidirectional network ring carrying data packets; the slave
arbiter: having a first ring port in data communication with a
first incoming link and a first outgoing link and a second ring
port in data communication with a second incoming link and a second
outgoing link connected to the bidirectional network ring;
operating to receive HELLO messages on the first incoming link and
on the second incoming link and to send HELLO messages on the first
outgoing link and the second outgoing link; maintaining an imposed
virtual break on either the first ring port or the second ring port
to block the passage of user data packets when the bidirectional
network ring is fully functional so as to prevent a data loop; and
reacting to the detection of a link failure on the second incoming
link by: imposing a break on the second incoming link so that user
data packets are blocked from passing into the slave arbiter but
the imposed break does not block any incoming HELLO messages;
imposing a break on the second outgoing link so that user data
packets are blocked from leaving the slave arbiter on the second
outgoing link; removing the imposed virtual break from either the
first ring port or the second ring port; and detecting the
cessation of the link failure on the second incoming link; then
imposing a virtual break on either the first ring port or the
second ring port before removing the imposed break on the second
incoming link and the imposed break on the second outgoing
link.
15. The slave arbiter of claim 14 wherein the link failure detected
on the second incoming link was indicated by an OPER DOWN.
16. The slave arbiter of claim 14 wherein the link failure detected
on the second incoming link was indicated by a gap in the receipt
of HELLO messages beyond a set threshold.
17. A slave arbiter for use in a bidirectional network ring for a
bidirectional network ring that is manipulated to provide a virtual
break in order to operate within a network requiring a tree
topology, the bidirectional network ring carrying data packets; the
slave arbiter: having a first ring port in data communication with
a first incoming link and a first outgoing link and a second ring
port in data communication with a second incoming link and a second
outgoing link connected to the bidirectional network ring;
operating to receive HELLO messages on the first incoming link and
on the second incoming link and to send HELLO messages on the first
outgoing link and the second outgoing link; maintaining an imposed
virtual break on either the first ring port or the second ring port
to block the passage of user data packets when the bidirectional
network ring is fully functional so as to prevent a data loop; and
after receiving a control signal on the second incoming link
indicating that another node on the bidirectional network ring is
imposing a bidirectional break, reacting to the received control
signal by: removing the imposed virtual bleak from either the first
ring port or the second ring port; and detecting that the node on
the bidirectional network ring that imposed the bidirectional break
will soon remove the bidirectional break and responding by imposing
a virtual break on either the first ring port or the second ring
port.
18. The slave arbiter of claim 17 wherein: the slave arbiter
receives the control signal on the second incoming link indicating
that another node on the bidirectional network ring is imposing a
bidirectional break by discerning between a first type of HELLO
message that indicates a lack of network problems over the path
traveled by the HELLO signal and a second type of HELLO message
that conveys the control signal indicating that a node on the
bidirectional network ring is imposing a bidirectional break; and
the slave arbiter detects that the node on the bidirectional
network ring that imposed the bidirectional break will soon remove
the bidirectional break by detecting the resumption of receipt of
the first type of HELLO message on the second incoming link.
19. The slave arbiter of claim 17 wherein: the slave arbiter
receives the control signal on the second incoming link indicating
that another node on the bidirectional network ring is imposing a
bidirectional break in a separate first type of control message not
part of a HELLO message; and the slave arbiter detects that the
node on the bidirectional network ring that imposed the
bidirectional break will soon remove the bidirectional break by
detecting the cessation of receipt of the first type of control
message on the second incoming link.
20. The slave arbiter of claim 17 wherein: the slave arbiter
receives the control signal on the second incoming link indicating
that another node on the bidirectional network ring is imposing a
bidirectional break in a separate first type of control message not
part of a HELLO message; and the slave arbiter detects that the
node on the bidirectional network ring that imposed the
bidirectional break will soon remove the bidirectional break by
detecting the receipt of a second type of control message on the
second incoming link.
21. A ring device for use in a bidirectional network ring for a
bidirectional network ring that is manipulated to provide a virtual
break in order to operate within a network requiring a tree
topology, the bidirectional network ring carrying data packets; the
ring device: having a first ring port in data communication with a
first incoming link and a first outgoing link and a second ring
port in data communication with a second incoming link and a second
outgoing link connected to the bidirectional network ring;
operating to send a first type of HELLO message onto the first
outgoing link and the second outgoing link, the receipt of the
first type of HELLO messages on a remote device on the network ring
indicating the functioning of the network between the ring device
and the remote device; and reacting to the discernment of a link
failure on the second incoming link by: imposing a break on the
second incoming link so that user data packets are blocked from
passing into the ring device but the imposed break does not block
any incoming HELLO messages; imposing a break on the second
outgoing link so that user data packets are blocked from leaving
the ring device on the second outgoing link; sending a type of
HELLO message and a control signal out the second outgoing link
such that the adjacent node receives a HELLO message to provide
assurance that the second outgoing link from that ring device is
operating and to convey to the remote device, even if the remote
device is not the adjacent node, the control signal that the ring
device is imposing a bidirectional break so that the virtual break
should be removed; detecting the cessation of the link failure on
the second incoming link; and delaying the removal of the imposed
break for user data packets on the second outgoing link and on the
second incoming link for a time period sufficient to allow the
bidirectional network ring to impose a virtual break at the remote
device.
22. The ring device of claim 21 wherein the link failure is the
detection of a problem with the communication signal.
23. The ring device of claim 22 wherein the link failure is
discerned by detecting a gap in the receipt of a HELLO message on
the second incoming link for a period in excess of a specified
value.
24. The ring device of claim 22 wherein the operation of sending
the type of HELLO message and the control signal out the second
outgoing link such that the adjacent node receives a HELLO message
to provide assurance that the second outgoing link from that ring
device is operating and to convey to the remote device, even if the
remote device is not the adjacent node, the control signal that the
ring device is imposing a bidirectional break so that the virtual
break should be removed is achieved by sending a second type of
HELLO message that differs from the first type of HELLO message in
that the second type of HELLO message indicates that the ring
device is imposing a bidirectional break.
25. The ring device of claim 24 wherein the ring device is adapted
to respond to the receipt of a second type of HELLO message on the
first incoming link by sending a second type of HELLO message on
the first outgoing link.
26. The ring device of claim 22 wherein the operation of sending
the type of HELLO message and the control signal out the second
outgoing link such that the adjacent node receives a HELLO message
to provide assurance that the second outgoing link from that ring
device is operating and to convey to the remote device, even if the
remote device is not the adjacent node, the control signal that the
ring device is imposing a bidirectional break so that the virtual
break should be removed is achieved by sending a HELLO message and
a separate control message outside of a HELLO message that
indicates that the ring device is imposing a bidirectional
break.
27. A network ring that is manipulated to provide a virtual break
in order to operate within a network requiring a tree topology, the
network carrying data packets, the network ring comprising: A) a
Master Arbiter (MA) comprising: a MA first port; with a first MA
incoming link and a first MA outgoing link; and a MA second port;
with a second MA incoming link and a second MA outgoing link; B) a
Slave Arbiter (SA) comprising: a set of SA ports including: a SA
first port normally connected to the MA first port by a full duplex
communication connection; a SA second port normally connected to
the MA second port by a full duplex communication connection; and a
SA control system, the SA control system comprising controls over
the introduction and removal of a virtual break to prevent the
formation of a communication ring for data packets within the
network requiring a tree topology through the use of at least one
non-forwarding state wherein control traffic is forwarded but data
packets are not forwarded; the SA control system comprising: at
least one rule causing the operation of one SA Port in a
slave-forwarding state wherein the SA port forwards data packets
and control traffic received at the SA port; at least one rule
causing the operation of one SA Port in one of at least one
non-forwarding states wherein control traffic is forwarded but data
packets are not forwarded; at least one rule causing the Slave
Arbiter to operate when the ring is fully operational such that one
SA Port is in a slave-forwarding state and the other SA Port is in
a non-forwarding state wherein control traffic is forwarded but
data packets are not forwarded; at least one rule for changing the
state of the Slave Arbiter to have both the SA first port and the
SA second port acting to forward data packets after receipt of a
control signal indicating that a device other than the Slave
Arbiter is imposing a bidirectional break in the network ring; and
at least one rule for changing the state of one of the two SA Ports
to a non-forwarding state wherein control traffic is forwarded but
data packets are not forwarded after discerning that the device
other than the Slave Arbiter is going to remove the imposed
bidirectional break in the network ring.
28. The network ring of claim 27 wherein the control signal
received by the Slave Arbiter indicating that a device other than a
Slave Arbiter is imposing a bidirectional break is received in a
HELLO message that conveys to the Slave Arbiter that the incoming
link to the Slave Arbiter that received the HELLO message is
working between the Slave Arbiter and an adjacent network device,
which may be the Master Arbiter.
29. The network ring of claim 27 wherein the SA control system
further comprises: at least one rule for detecting a time gap
between receipt of HELLO messages in excess of a set value and
responding to that detected time gap by imposing a bidirectional
break to block data packets but not HELLO messages on both the
incoming link with the detected time gap and the outgoing link on
that same SA port then changing the state of the Slave Arbiter to
have both the SA first port and the SA second port acting to
forward data packets; and at least one rule for changing the state
of one of the SA Ports to a non-forwarding state wherein control
traffic is forwarded but data packets are not forwarded after
discerning the receipt of a HELLO message on the incoming link with
the detected time gap before removing the imposed bidirectional
break.
30. The network ring of claim 27 wherein the SA control system
further comprises: at least one rule for detecting a problem on
either the first incoming link or the second incoming link and
responding to that detected problem by imposing a bidirectional
break to block data packets but not HELLO messages on both the
incoming link with the detected problem and the outgoing link on
that same SA port then changing the state of the Slave Arbiter to
have both the SA first port and the SA second port acting to
forward data packets; and at least one rule for changing the state
of one of the SA Ports to a non-forwarding state wherein control
traffic is forwarded but data packets are not forwarded after
discerning the cessation of the detected problem on the incoming
link with the detected problem before removing the imposed
bidirectional break.
31. The network ring of claim 27 further comprising a ring relay
(RR) positioned between the MA first port and the SA first port,
the ring relay comprised of: a RR MA-side port with a RR MA
incoming link and a RR MA outgoing link; and a RR SA-side port with
a RR SA incoming link and a RR SA outgoing link; such that the full
duplex communication between the SA first port and the MA first
port runs through the ring relay and the ring relay operates after
detecting a problem with the RR-MA incoming link to send out a
control signal on both the RR MA-side port and on the RR SA-side
port and acts to impose a bidirectional break on both the RR MA
incoming link and the RR MA outgoing link.
32. The network ring of claim 27 further including: a ring relay
(RR) positioned between the MA first port and the SA first port,
the ring relay comprised of: A RR MA-side port with a RR MA
incoming link and a RR MA outgoing link; and A RR SA-side port with
a RR SA incoming link and a RR SA outgoing link; such that the full
duplex communication between the SA first port and the MA first
port runs through the ring relay and the ring relay operates after
detecting a problem with the RR-MA incoming link to send out a
control signal on the RR MA-outgoing link and imposing a
bidirectional break on both the RR MA incoming link and the RR MA
outgoing link; and the Master Arbiter further including reacting to
the receipt of the control signal on the first MA incoming link by
sending a control signal out on the first MA outgoing link so that
the control signal can be passed through the ring relay towards the
Slave Arbiter.
33. The network ring of claim 32 wherein control signal indicating
that a device other than the Slave Arbiter is imposing a
bidirectional break in the network ring is conveyed to the Slave
Arbiter in a second type of HELLO message that is different from
the first type of HELLO message used when not sending the control
signal that a device other than the Slave Arbiter is imposing a
bidirectional break in the network ring; and after a HELLO message
of the second type is sent from the ring relay to the Master
Arbiter, HELLO message of the second type is sent from the Master
Arbiter to the ring relay, and a HELLO message of the second type
is transmitted from the ring relay toward the Slave Arbiter.
34. The network ring of claim 33 wherein: the ring relay blocks
HELLO messages of the first type from passing from the RR SA
incoming link to the RR MA outgoing link; and the Master Arbiter
creates a new HELLO message of the second type for transmission on
the first MA outgoing link and the ring relay passes the new HELLO
message of the second type from the RR MA incoming link to the RR
SA outgoing link.
35. A network device comprising: a pair of access ports consisting
of: a first port with a first incoming link and a first outgoing
link; a second port with a second incoming link and a second
outgoing link; at least one user port; and a control system for the
network device; the control system adapted to allow this network
device to provide redundant access for the at least one user port
to an existing network through a first access device connected to
the first port and a second access device in data communication
with the first access device and connected to the second port such
that a HELLO message transmitted out the first outgoing link can
travel via the first access device then the second access device to
the second incoming link and a HELLO message transmitted out the
second outgoing link can travel via the second access device then
the first access device to the first incoming link; the control
system including: at least one rule to impose a virtual break on
either the first port or the second port when HELLO messages are
being received on both access ports, the virtual break blocking
data packets but not control messages such as HELLO messages; at
least one rule to detect and react to a HELLO timeout on both
access ports by removing the virtual break from either the first
port or the second port so that the at least one user port can have
access to the existing network through the first access device
connected to the first port if that is possible or through the
second access device connected to the second port if that is
possible; at least one rule to detect and react to a HELLO timeout
on one access port and an OPER DOWN on the other access port by
removing the virtual break from either the first port or the second
port so that the at least one user port can have access to the
existing network through the first access device connected to the
first port if that is possible or through the second access device
connected to the second port if that is possible, this at least one
rule not necessarily distinct from the rule to detect and react to
a HELLO timeout on both access ports; at least one rule to detect a
unidirectional break having an OPER DOWN or a HELLO timeout on one
access port but not the other access port and to react by imposing
a bidirectional break on the incoming link and the outgoing link on
the access port with incoming link experiencing an OPER DOWN or a
HELLO timeout before removing the virtual break from the access
port with the virtual break; and at least one rule to detect and
react to the cessation of the unidirectional break having an OPER
DOWN or a HELLO timeout by imposing a virtual break on one access
port before removing bidirectional break.
36. The network device of claim 35 wherein the control system uses
a delay timer to delay the removal of a virtual break a time period
sufficient to allow the imposition of the bidirectional break so
that a data loop is not created when the virtual break is
removed.
37. A method for operating a network ring that is manipulated to
provide a virtual break in order to operate within a network
requiring a tree topology, the network ring carrying data packets;
the method comprising: imposing a virtual break at a first network
device, the first network device comprised of a first port with a
first incoming link and a first outgoing link and a second port
with a first incoming link and a second outgoing link, the virtual
break imposed whenever the first network device is receiving HELLO
messages of a first type on both the first incoming link and the
second outgoing Link; removing the virtual break at the first
network device upon receipt of a control signal from another device
on the network indicating that the another device is imposing a
bidirectional break; and discerning that the another network device
is going to stop imposing the bidirectional break and imposing a
virtual break at the first network device.
38. The method of claim 37 wherein: the step of discerning that the
another network device is going to stop imposing the bidirectional
break includes discerning that the another network device is going
to stop imposing the bidirectional break upon receipt of a control
signal from the first network device that the first network device
has imposed the virtual break; and the step of imposing a virtual
break includes sending a control signal to the another network
device after the imposition of the virtual break to trigger the
removal of the bidirectional break.
39. The method of claim 37 wherein: the another network device is
comprised of a first port with a first incoming link and a first
outgoing link and a second port with a second incoming link and a
second outgoing link and the another device imposes a bidirectional
break and sends the control signal to the first network device when
the another device ceases to receive HELLO messages of any type for
more than a prescribed period on the first incoming link; and the
another network device responds to the resumption of receipt of
HELLO messages on the first incoming link: by indicating to the
first network device that the another network device will stop
imposing the bidirectional break so that the first network device
can impose a virtual break before the bidirectional break is
removed; and by removing the bidirectional break.
40. The method of claim 37 wherein the control signal is sent to
the first network device in a HELLO message of a second type
different from the HELLO message of the first type.
41. The method of claim 40 wherein the HELLO message of the second
type sent by the another network device are received by a third
network device between the first network device and the another
network device and the receipt of the HELLO message of the second
type on a first incoming link for the third device precludes the
third network device from reacting on a lack of HELLO messages of
any type on the first incoming link for more than a prescribed
period.
42. The method of claim 40 wherein a third network device receives
the HELLO message of the second type from the another network
device and responds by sending a HELLO message of the second type
back towards the another network device which in turn sends the
HELLO message of the second type away from the third network device
and towards the first network device.
43. The method of claim 37 wherein the control signal is sent to
the first network device in a control message that is not a type of
HELLO message.
44. The method of claim 37 wherein the first outgoing link of the
first port of the first network device is connected to a second
incoming link on a second port on an adjacent network node which is
the another network device, and a second outgoing link on the
second port on the another network device is connected to the first
incoming link on the first network device and the response to a
bidirectional break between the first port of the first network
device and the second port of the another network device is the
imposition of bidirectional breaks at both the first port of the
first network device and at the second port of the another network
device and the removal of the virtual break at the first network
device.
Description
BACKGROUND OF THE INVENTION
Field of the Invention
This invention relates to communication networks, and more
particularly, to an apparatus and method for Ethernet equipment in
a ring topology.
As Ethernet is deployed in Metro and Access networks, and services
are offered on these networks, there is a desire to maintain
SONET-like resiliency (i.e. recover from a fault in less than 50
ms). One common means of providing resiliency is with a ring
topology. However, Ethernet switches will not work properly if
there is a ring or loop in the topology. Protocols such as IEEE
802.1d Spanning Tree Protocol (STP) or IEEE 802.1w Rapid
Reconfiguration were invented to detect and remove loops. However,
they are slow and cannot achieve path restoral within 50 ms as
desired.
To solve this problem, the IEEE is working on 802.17 Resilient
Packet Ring (RPR). Others are looking at Multiprotocol Label
Switching (MPLS) with Fast Reroute capabilities. Both of these
approaches are quite complex. RPR requires a new Media Access
Control (MAC) Layer, and MPLS requires extensive signaling. Because
of the complexities, these approaches will drive up the cost of the
nodes on the ring.
This invention introduces a new way (Protected Switching Ring or
"PSR") of providing protection for Ethernet deployed in a ring
topology with resiliency that does not require a new MAC layer, and
that can be built using low cost Ethernet chips and methods.
This invention differs from some previous inventions. One of
interest is described in U.S. Pat. No. 6,430,151, granted on Aug.
6, 2002. PSR is similar to '151 in that: Both are based on nodes
arranged in a ring topology. Both aim to overcome the limitations
of STP. Both describe making or breaking a ring based on the
passage or blockage of test messages. Both have two classes of
nodes on the ring, one of which is a controller or Master.
Some of the differences between PSR and the teachings of the '151
include: '151 is composed of bridging nodes that do dynamic layer 2
learning, while PSR is composed of nodes that are configured to
switch (add and/or drop) packets based on a VLAN tag. '151 patent
has a single redundancy manager (RM), while PSR can support dual
redundancy Ring Arbiters (RA). PSR can operate in the presence of a
failed RA, thus providing a higher level of availability. The nodes
in the '151 patent learn an association between ports and MAC
addresses for ring traffic. When the topology changes, the RM of
the '151 patent must send a "flush" message to tell the nodes to
clear their databases. In contrast, the Ring Relay ("RR") nodes in
PSR always send messages (both data and control) around the ring in
both directions, thus removing half of the propagation delay from
the recovery time. In this way a flush command is not needed to
redirect traffic on the ring, thus reducing the recovery time. '151
patent can cause packets to be duplicated during a restoral as
there will be a ring upon restoral. Duplication of packets violates
the IEEE 802.3 specifications. The state machines in PSR prevent
this. Since nodes in PSR are not performing learning for ring
traffic, there is less overhead and a higher packet rate can be
sustained for a given amount of processing power.
Another approach to the problem is described in U.S. Pat. No.
4,354,267. The '267 patent describes a set of homogeneous layer 2
devices arranged in a ring. Each node in the '267 patent forwards
packets around the ring, and the originator removes the packet.
Some of the differences between PSR and the teachings of the '267
patent include: '267 patent assumes that data sent that is sent one
way around the ring makes it all the way around. In layer 2
systems, each node may pick off packets addressed to it, so this
assumption is not valid. '267 patent assumes that each node can
repair a fault. See claim 1 in column 10, starting at line 34, and
claim 5, in column 12, starting at line 38. In contrast, PSR
concentrates the recovery mechanism in just two nodes.
SUMMARY
Normal 802.3 Ethernet requires a tree topology. If a ring or a loop
exists, then packets will be forwarded around the ring
indefinitely. STP was created to solve this problem by detecting
and breaking any rings. If the ring is broken, then there is no
possibility of-packets being propagated forever.
This invention shows how to virtually break the ring such that all
nodes can communicate with each other, and how to remove the
virtual break when a real failure occurs. This is accomplished by
placing intelligent nodes on the ring that work together to
virtually break and restore the ring.
In PSR, the nodes communicate between and among themselves to
determine when and where a break occurs. The relevant state
machines for a preferred embodiment of the present invention are
contained within this disclosure.
This application extends and expands on the ability of PSR to
handle unidirectional breaks in various ring topologies.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows an example of prior art.
FIG. 2 shows an example Protected Switching Ring in the Full Ring
mode in normal operation.
FIG. 3 shows an example Protected Switching Ring in the Full Ring
mode during a failure.
FIG. 4 shows an example Protected Switching Ring in the High
Availability mode in normal operation.
FIG. 5 shows an example Protected Switching Ring in the High
Availability mode during a failure.
FIG. 6 shows the state machine for a Ring Arbiter node in the Full
Ring mode.
FIG. 7 shows the state machine for a Ring Relay node in the Full
Ring mode.
FIG. 8 shows the state machine for the Ring Side of a Ring Arbiter
node in the High Availability mode.
FIG. 9 shows the state machine for the Extension Side of a Ring
Arbiter node in the High Availability mode.
FIG. 10 illustrates a unidirectional ring break.
FIG. 11 shows the "Dual Homing" embodiment providing User Ports
1140 with redundant links to the existing network.
FIG. 12 shows a "Dual Homing" embodiment providing User Ports 1231,
1232, and 1233 with redundant links to the existing network and
showing the individual unidirectional links rather than
bidirectional links in order to discuss a method of responding to a
unidirectional break.
DETAILED DESCRIPTION
The present invention will now be described more fully hereinafter
with reference to the accompanying drawings, in order to disclose
selected embodiments. This invention may, however, be embodied in
many different forms and should not be construed as limited to the
embodiments set forth herein; rather, these embodiments are
provided so that this disclosure will be thorough and complete, and
will fully convey the scope of the invention to those skilled in
the art. Like numbers refer to like elements throughout.
Overview
The Protected Switching Ring (PSR) enables building of partial or
full ring topologies from low-cost Ethernet equipment, while
providing for sub-50 millisecond recovery from equipment or link
faults. PSR nodes support the transport of point-to-point
port-switched connections across the ring topology. During normal,
non-fault operation, one port in the ring will be blocked to user
traffic, thus preventing a loop. In the event of a fault in the
ring, the blocked port will be unblocked, allowing access to all
nodes on the ring.
Two topologies using the present invention are described below. The
first topology is the PSR Full Ring ("FR") configuration that
consists of a full ring of PSR nodes. Port-switch connections can
be configured between any two subscriber ports on the ring. The
second topology is the PSR High-Availability (HA) configuration.
This configuration provides a partial-ring extension of a SONET or
RPR ring, or a partial-ring addition to existing layer-2 switching
equipment. In either case, a path is engineered through the
existing equipment to complete the path for the PSR protocol
traffic and user data.
PSR nodes are designated as Ring Arbiters or Relays. Each ring
contains two Ring Arbiters. The Ring Arbiters communicate with a
"hello" protocol to coordinate the blocking or forwarding of user
traffic. In a preferred embodiment, the PSR Ring Arbiter ports take
on the role of Master or Slave on the ring according to their
relative node priority. In a preferred embodiment the priority
could be a unique identifier, such as a MAC address. In a highly
preferred embodiment, the priority can be the concatenation of an
operator-configurable priority with the MAC address (or other
unique identifier) such that the priorities of two nodes would
never be equal. In either of these preferred embodiments, the
reception of a HELLO with the same priority would indicate a ring
with only one arbiter, where that arbiter was receiving its own
HELLOs.
In general, during normal fault-free operation of two Ring
Arbiters, the Slave Ring Arbiter will block one of its ring ports
in order to terminate the ring loop. A ring may contain one or more
Relay nodes. The Relay nodes may be distributed in any fashion
around the ring, although some benefit is provided by distributing
approximately equal numbers of Relays on each "side" of a full-ring
configuration.
In addition to the "hello" protocol, each node performs a
"discovery" protocol that allows each node to know about all the
other nodes on the ring. The discovery protocol is also used to
detect persistent ring faults and to generate the associated
alarms. Both protocols operate at layer 2, employing reserved
multi-cast MAC addresses.
IP connectivity between all ring nodes is accomplished over a
control VLAN used only for that purpose. This allows Telnet and a
UDP-based signaling protocol to operate between any nodes on the
ring. (An explanation of Telnet is not critical to the
understanding of the present invention but Telnet is a terminal
emulation program used with TCP/IP networks that allows remote
entry of commands that are treated as if input at the network
device.) Bridging techniques are used to provide the connectivity
for these IP-based applications; all user traffic is transported
across the ring using port-switching. As such, all user traffic is
point-to-point across the ring; traffic from a subscriber-port/VLAN
on one node is connected to a subscriber-port/VLAN on another
node.
An additional embodiment of the present invention addresses a
partial failure of a network link so that the communication link is
lost in only one direction. Yet, another embodiment uses a single
arbiter to provide a high reliability connection of user ports to
an existing network ring by creating a switching ring with the
arbiter and two network ring access points.
Ring Nomenclature
When the PSR is configured, two ports are designated as the ring
ports and may be called east and west ports. Also the node type is
given to distinguish Ring Arbiter types and Relays (also called
Ring Relays or Relay Nodes). The Ring Arbiter type may be
High-Availability (HA) or Full-Ring (FR). The two Ring Arbiters on
the ring must be of the same type. When speaking of a specific ring
port, the partner port refers to the other port of the pair of ring
ports on that Ring Arbiter or Ring Relay.
An additional distinction is made in the case of a HA Ring Arbiter.
The port of the HA Ring Arbiter connected to the existing SONET or
RPR ring is designated the "extension side" (ES) port. This port
interfaces with the existing equipment for which we wish to extend
a ring segment. The other Ring Arbiter port is referred to as the
"ring side" port. It is connected to a string of one or more Ring
Relays or directly to the other Ring Arbiter.
HELLO Protocol
Each PSR Ring Arbiter periodically issues a "HELLO" protocol packet
out each ring port. In a preferred embodiment each PSR Ring Arbiter
issues a "HELLO" protocol packet out each ring port every 10
milliseconds. The packet uses a special multicast MAC address as
the destination address. The Relay nodes are configured to have the
data plane pass the packet from one ring port to the other, so a
Relay node adds only a small amount of delay as the packet moves
from one Ring Arbiter to the other. The remote Ring Arbiter node
will terminate the packet and send the packet to the control plane.
The control plane uses the presence of the new packet and some
control information to drive its state machine for the Ring Arbiter
ports. The absence of a new HELLO message for 30 milliseconds
constitutes a ring timeout. If the timeout persists for 1.5
seconds, a ringfailure is declared and the appropriate alarm is
issued.
The significantly longer period used as a trigger for a ring
failure keeps a short intermittent problem from being deemed ring
failures though the problems may be handled by the declaration of
ring timeouts. In one embodiment, the ring failure is detected by
loss of Discovery messages, described below. One of skill in the
art could implement the ring failure to be based on the absence of
HELLO messages rather than Discovery messages. One of skill in the
art would appreciate that the HELLOs are not processed at the RR
nodes, whereas the Discovery messages are. HELLOs therefore
propagate around the ring faster than Discovery messages. A timeout
threshold for loss of HELLOs can be set lower than an equivalent
threshold for Discovery messages.
A ring timeout causes the state machines to transition a Slave Ring
Arbiter port to a FORWARDING state. This response ensures that any
loss in connectivity due to a single failure across the ring will
only persist for 50 milliseconds or less.
In a preferred embodiment the sequence number in the HELLO PDU is
used at the receiving Ring Arbiter to distinguish the arrival of a
new HELLO PDU. Those of skill in the art will recognize that other
methods could be employed to detect the arrival of a new HELLO PDU.
The Relay nodes do not process the HELLO PDUs; they only forward
them between ring ports.
Discovery Protocol
The discovery protocol is an optional protocol that can be
implemented in order to add functionality. Note since the discovery
protocol is not a necessary requirement of the state machines for
any of the Ring Arbiters, Protected Switching Rings in accordance
with the teachings of the present invention could be implemented
without implementing the discovery protocol.
The discovery protocol also uses a special multicast destination
MAC, but runs every 500 milliseconds. The discovery PDU is
originated by the Ring Arbiters, appended to by intervening Relay
nodes, and terminated at the remote Ring Arbiter. As the discovery
PDU traverses the path between Ring Arbiters, each node in the path
appends its management IP address, egress port for the PDU, and
node type to the PDU. Since the discovery messages are flowing in
both directions on the ring, each node on the ring can see the path
of nodes to each Ring Arbiter on each of its ring ports. For
example in FIG. 2, the Ring port 210 will receive a discovery
message on one port directly from the RA 200 and will receive the
other discovery message from the RA 225 after that discovery
message passes through the ring port 220. Thus after receiving the
two discovery PDUs, each ring port knows the identity of all
devices between the ring port and each RA.
Additionally, as each Ring Arbiter constructs the discovery message
to send out a ring port, the Ring Arbiter adds the completed node
list received at its partner port. This allows every node in the
PSR to know all the IP addresses of the nodes in the ring.
In the event of a ring or node failure, the Relay nodes closest to
the point of failure will originate the discovery message. In other
words, if a relay fails to receive a discovery PDU from its
upstream neighbor (due to a link or node failure), then the relay
will create and send a discovery PDU in the downstream direction.
All downstream nodes will detect that the Ring Arbiter is no longer
the originator of the discovery message and declare a fault alarm.
If a node either does not receive a Discovery message or receives a
Discovery message without a Ring Arbiter as the originator, a ring
failure is declared after 1.5 seconds. The fault is cleared when
the node receives a Discovery message with a Ring Arbiter as the
originator.
PSR Data Plane for User Traffic
User traffic may enter and leave the PSR at any Ring Arbiter or
relay node. A PSR connection defines the entry and exit points for
a full-duplex flow of user traffic across the ring. This flow is
defined by a pair of port/VLAN ID/PSR Node Address tuples. The
connection defines a path through the ring between 2 user ports,
each residing on a PSR node, configured to carry the user traffic
for specific or all VLAN IDs on that port.
As the user traffic enters the ring, a ring tag is added to the
packet. The ring tag is a VLAN tag and is unique on the ring. The
ring tag defines a given connection between two ring nodes. At the
egress node of the PSR connection, the ring tag is removed from the
frame before forwarding to the user port. In this way, the VLAN
tags present in the user data are transparently transported across
the ring. VLAN IDs used on one user port do not interfere with IDs
used on another user port.
A PSR node is either an endpoint of a given connection or a transit
node for that connection. A PSR node is an endpoint for a
connection if one of its user ports is specified in the definition
of the given connection. The node is a transit node if neither
endpoint of the connection resides on the node. In either case, a
switch table used by the data plane is configured on each PSR node
to either terminate one end of a given connection or to act as a
transit node for that connection. When a node is a transit node for
a given connection, the node simply transfers frames from one ring
port to the other, based on the ring tag, without modification.
When a node is an endpoint node for a given connection, the data
plane directs the data arriving on a ring port to the correct user
port and removes the ring tag. Conversely, the node's data plane
directs user packets from the given user port with the given VLAN
ID to the ring ports, adding the correct ring tag in the
process.
PSR Control Plane for Control Traffic
A PSR requires a mechanism to transport HELLO PDUs, discovery PDUs,
and IP traffic for ring control applications between PSR nodes.
While user traffic transport is transported using switching
techniques, in a preferred embodiment the control functions are
transported using bridging techniques. By using bridging
techniques, full PSR node control connectivity is attained with all
nodes appearing on the same IP subnet. This makes configuration
much simpler.
One ring tag is reserved for PSR control traffic. The data plane
uses learning procedures and forwarding table lookups to direct
control traffic to the correct PSR node. Note that the use of
learning procedures and forwarding table lookups for the direct
control traffic imposes an overhead burden that is orders of
magnitude smaller than the overhead needed to use learning
techniques for user data traffic. In the preferred embodiment the
HELLO and discovery messages use known multicast MAC addresses and
thus do not add additional learned database entries to be
forwarded. Flushing is not needed for the control traffic upon
failure, recovery, or reconfiguration of the ring, as the new port
entries are learned from bidirectional traffic after a short period
of time.
While the use of bridging for control traffic is preferred, it is
not required in order to implement the present invention. The
present invention could be implemented to use switching techniques
for data packets and some or all types of control traffic. Care
must be taken in creating this variation that the control traffic
described in this application as passing when data packets are
blocked, must be allowed to pass.
Example Recovery for Full Ring
Fault Detection
FIG. 3 shows a full ring where the link 1325 fails between nodes
310 and 320. This means that RA nodes 300 and 325 are unable to
communicate with each other via the left hand side of the ring.
Prior to the failure, assume that RA node 325, the Slave Ring
Arbiter, is blocking traffic on link 1330 (thus no counterclockwise
communication on 1330) and forwarding traffic on link 1335. Also,
any user traffic arriving on link 1330 is discarded. So clockwise
traffic on 1330 is discarded at the 1330 side of RA 325.
Communications to subscriber ports connected to RA 325 reach those
ports through counterclockwise communication over link 1335 to RA
325.
Assuming RA node 300 was the Master Ring Arbiter, when RA node 325
detects the loss of communication; RA node 325 will start
forwarding traffic to the right hand side of the ring onto link
1330 and accepting user traffic arriving on link 1330 and relaying
the traffic to link 1335 and to the subscriber ports of RA 325.
This will restore communications between all of the nodes on the
ring. At this point, RA 325 is forwarding traffic on both ring
ports. The ring port that is facing link 1335 is in MASTER
FORWARDING state, and the ring port that is facing link 1330 is in
SLAVE FORWARDING state.
Link Restoral
When link 1325 is restored, RA node 325 needs to block one of its
ring ports to prevent a loop in the ring. When RA node 325 receives
the first HELLO on link 1335 (due to the restoration of link 1325),
RA node 325 will see that the partner port to the port that is
facing link 1335 is in SLAVE FORWARDING state. RA node 325 will
move the port that is facing link 1335 to the BLOCKING state.
Assuming that the Ring ports of nodes 310 and 320 connected to link
1325 went to an OPER DOWN state during the failure, the TIMING
state in the relay nodes 310 and 320 will prevent forwarding of
traffic until the Slave Ring Arbiter has time to switch from MASTER
FORWARDING to BLOCKING on the 325 side of the Ring Arbiter. OPER
DOWN is an indication from the physical or transport layer that a
link is not operational. It is normally based on the detection of
loss or corruption of the incoming electrical or optical
signal.
The advance to the TIMING state is triggered by the reception of a
HELLO message. This TIMING state delay in the resumption of
operation of relay nodes 310 and 320 prevents duplicate packets
from reaching a given destination when the failed link is restored.
To illustrate the value of this delay in the Ring Relay ports,
consider a message coming to Ring Relay 305 to a subscriber port
connected with Ring Relay 310 just before link 1325 is restored.
Ring Relay 305 operating normally will send the same message onto
link 1300 and link 1320. The message traveling counterclockwise
reaches Ring Relay 310 in a conventional way. The message traveling
clockwise to Ring Relay 310 will pass through Ring Arbiter 325 onto
link 1335 as the west Port is operating in MASTER FORWARDING. When
link 1325 is restored, there is a path for a duplicate message to
cross link 1325 to Ring Relay 310. This potential is eliminated if
the Ring Relay ports undergo a suitable delay between receipt of
the first HELLO and the ultimate state of FORWARDING. Note that the
HELLO messages from Ring Arbiter 300 to Ring Arbiter 325 and from
Ring Arbiter 325 to Ring Arbiter 300 will pass over link 1325 as
soon as it is restored as the HELLO messages are not blocked at any
port in any state.
The preferred embodiments disclose using a timing delay to ensure
that a port progressing from OPER DOWN to operational delays
sending data packets long enough for the Slave arbiter to impose a
virtual break. One of skill in the art will recognize that the use
of the timer could be replaced by a control signal sent by the
Slave arbiter after it has successfully imposed the virtual break.
In either case, the port does not go to fully operational until
after the virtual break has been imposed to preclude the creation
of a ring for data packets.
Example Recovery for HA Ring
Fault Detection
FIG. 4 shows a HA ring under normal fault-free operation. The ES
Slave port 1440 is in the BLOCKING state to prevent a ring loop.
FIG. 5 shows a HA ring where the link 1520, between nodes 510 and
520, fails. As for the full ring case, the bidirectional failure of
link 1520 means that the Ring Arbiter nodes 500 and 530 are unable
to communicate over the left side (Ring Side) portion of the HA
ring. Assuming Ring Arbiter node 530 is the Slave, its ES port (the
facing link 1540) would be un-blocked when the failure is detected.
Fault detection and subsequent un-blocking of the Slave Ring
Arbiter port is fundamentally the same as for the full ring case
described above.
Link Restoral
In a preferred embodiment, the HA ring favors the Ring Side once
the fault is removed. Instead of leaving the Slave Ring Arbiter ES
port (the port facing link 1540) in the forwarding state and
blocking the Ring Side port (the port facing link 1530), the HA
Slave Ring Arbiter 530 always returns to a FORWARDING state on the
Ring Side segment and blocks the ES port.
The Ring Side segment of the HA ring is favored in order to
minimize HA ring traffic on the existing SONET or RPR ring as this
will cut some of the user traffic on the SONET ring segment between
the Ring Access Equipment as one direction will be blocked (thus
counterclockwise traffic from port 1440 will be blocked while
clockwise traffic from 1400 will continue to travel on the SONET
Ring.
Nomenclature for State Diagrams
In the following descriptions, "isMaster" is based on the last
received HELLO. If no HELLO was ever received on the port, then
isMaster is based on the partner's last HELLO. If no HELLOs have
ever been received by either port, then isMaster is "true". The
Boolean variable "isSlave" is the logical negation
of"isMaster".
The term "PartnerHelloTimeout" indicates that the partner port's
age timer has timed out. The designation "RxHello<Node" means a
HELLO message has been received whose priority is less than that of
the receiving node. This event would cause the receiving node to
consider itself a Master.
Full Ring Mode--Ring Arbiter
FIG. 6 shows the state diagram for a RA node. Each of the two ports
on an RA node runs a copy of this state machine.
Description of States
The state machine of FIG. 6 has the following states.
TABLE-US-00001 TABLE A Number State Description 600 PORT DOWN The
port is operationally down or has just been initialized. Entered
from any state. 610 BLOCKING The node is sending HELLOs, but not
forwarding data traffic. 620 SLAVE TIMING Node knows that it is a
Slave, but port is waiting for a timer to expire before moving to a
forwarding state. 630 MASTER TIMING Node knows that it is a Master,
but port is waiting for a timer to expire before moving to a
forwarding state. 640 SLAVE The port on a Slave Node is forwarding
packets FORWARDING 650 MASTER The port on a Master Node is
forwarding packets FORWARDING
Description of State Transitions
The table below describes the transitions of the state machine
shown in FIG. 6.
Note the fd timer reference below runs using a time that is a small
fraction of the time used for the age timers in the RA and Relay
nodes. This ensures that the relays are timed for a period long
enough after the expiration of the fd timer for the loop to be
broken. For example, one embodiment uses a 10 millisecond timer for
the RA and Relay nodes and the fd timer at just one "tick" (a
single 10 millisecond delay). This 10 millisecond interval is a
small fraction of the 30 millisecond interval used to detect a ring
timeout when a new HELLO message does not arrive within that
interval.
Note that the state machine for ring arbiters in the preferred
embodiment does not wait indefinitely to see a HELLO as long as the
ports of the ring arbiter are operationally UP. The goal is to let
the parts of the network ring operate even if other portions of the
network ring cannot operate.
TABLE-US-00002 TABLE B Num Event Action 1610 port operationally
down OR init block user traffic, cancel all timers 1615 port
operationally up start age timer 1620 age timer expires OR RxHello
< Node start fd timer 1625 fd timer expires restart age timer,
forward user traffic 1630 RxHello > Node start fd timer 1635 fd
timer expires AND partner not SLAVE restart age timer, FORWARD
forward user traffic 1640 age timer expires OR RxHello < Node
restart fd, age timer 1645 RxHello > Node restart fd timer 1650
Age timer expires OR RxHello < Node restart age timer 1655
RxHello > Node AND partner not SLAVE restart age timer FORWARD
1660 RxHello > Node AND partner SLAVE restart age timer
FORWARD
The following table shows the complete state transitions for the
Full-Ring Arbiter as shown in FIG. 6.
TABLE-US-00003 TABLE C Current State PORT SLAVE MASTER SLAVE MASTER
DOWN BLOCKED TIMING TIMING FORWARDING FORWARDING Event 600 610 620
630 640 650 Oper Down N/A PORT DOWN PORT DOWN PORT DOWN PORT DOWN
PORT DOWN Oper Up BLOCKED N/A N/A N/A N/A N/A Age Timer N/A MASTER
MASTER MASTER MASTER MASTER Expires TIMING TIMING TIMING FORWARDING
FORWARDING fd Timer Expires N/A N/A N/A MASTER N/A N/A FORWARDING
fd timer Expires N/A N/A SLAVE N/A N/A N/A AND Partner TIMING SLAVE
FORWARDING fd Timer Expires N/A N/A SLAVE N/A N/A N/A AND Partner
not FORWARDING SLAVE FORWARDING RxHello < Node N/A MASTER MASTER
MASTER MASTER MASTER TIMING TIMING TIMING FORWARDING FORWARDING
RxHello .gtoreq. Node N/A SLAVE SLAVE SLAVE N/A N/A TIMING TIMING
TIMING RxHello .gtoreq. Node N/A N/A N/A N/A N/A BLOCKED AND
Partner SLAVE FORWARDING RxHello .gtoreq. Node N/A N/A N/A N/A
SLAVE SLAVE AND Partner not FORWARDING FORWARDING SLAVE
FORWARDING
In a preferred embodiment, every 10 milliseconds, the two ports are
checked in the same order. The combination of variations in when
the HELLOs were generated plus transit delays may cause one HELLO
on one port to arrive before the other HELLO on the other port. In
any case, since one port is checked before the other then the
other, it always appears as though one HELLO arrives before the
other. The order that the ports are checked does affect which Slave
port is set to BLOCKING on the full ring.
One of skill in the art will recognize that any embodiment that
does not check one port before the other would need to address the
case of two HELLOs arriving essentially simultaneously at the two
ports.
TABLE-US-00004 TABLE D Time Port A Input State Change Port B Input
State Change 1 Port Down Port up 1615 to Port Down Port up 1615 to
Blocking Blocking HELLOs generated by other RA and sent towards
ports A and B of this RA. One HELLO arrives slightly before the
other. 2 Blocking RxHello > node 1630 to Slave Blocking Timing 3
Slave Timing Blocking RxHello > node 1630 to Slave Timing 4
Slave Timing fd timer 1635 to Slave Timing expires and SLAVE
partner not FORWARD SLAVE FORWARD 5 Slave forward Slave Timing
[cannot advance to Slave Forward as partner is in Slave Forward] 6
Link breaks 7 Slave Forward Link breaks, 1650 to Master Slave
Timing age timer Forwarding expires 8 Master Slave Timing fd timer
1635 to Forward expires and SLAVE partner not FORWARD SLAVE FORWARD
9 Master Slave Forward Forward 10 Link Restored 11 Master RxHello
> Node 1660 to Slave Forward Forward and Blocking partner Slave
Forward 12 Blocking RxHello > Node 1630 to Slave Slave Forward
Timing This continues until a port goes down, a link goes down, or
the node number of the other RA changes to become less than Node
(normally this would take a reconfiguration from the operator or
the substitution of another RA unit).
Full Ring Mode--Ring Relay
FIG. 7 shows the state machine for a Ring Relay node.
Description of States
The state machine of FIG. 7 has the following states.
TABLE-US-00005 TABLE E Number State Description 700 PORT DOWN The
port is operationally down or has just been initialized. Entered
from any state on an indication of the port going down due to a
loss of signal or other similar alarm. 710 AWAITING Port is
operationally up, but no HELLO HELLO has been received 720 TIMING
The port is waiting for the fd timer to expire 730 FORWARDING
Normal forwarding.
Description of State Transitions
The table below describes the transitions of the state machine
shown in FIG. 7.
TABLE-US-00006 TABLE F Number Event Action 1710 port operationally
down block user traffic, OR init cancel all timers 1715 port
operationally up start age timer 1720 age timer expires OR RxHello
start fd timer 1725 fd timer expires forward user traffic
High Availability Mode--Ring Arbiter--Ring Side
FIG. 8 shows the state machine for the Ring Sides (RS) of a Ring
Arbiter in HA mode.
Description of States
The state machine of FIG. 8 has the following states.
TABLE-US-00007 TABLE G Number State Description 800 PORT DOWN The
port is operationally down or has just been initialized. Entered
from any state. 810 BLOCKING The port is sending HELLOs, but is not
forwarding data traffic. 820 SLAVE FORWARDING The port on a Slave
Node is forwarding packets 830 MASTER The port on a Master Node is
FORWARDING forwarding packets
Description of State Transitions
The table below describes the transitions of the state machine
shown in FIG. 8.
TABLE-US-00008 TABLE H Number Event Action 1810 port operationally
down OR init block user traffic, cancel age timer 1815 port
operationally up start age timer 1820 (age timer expires AND
isMaster) Forward user traffic OR RxHello < Node 1825 (age timer
expires AND isSlave) OR Forward user traffic RxHello > Node 1830
RxHello < Node No action 1835 RxHello > Node No action
High Availability Mode--Ring Arbiter--Extension Side
FIG. 9 shows the state machine for the Extension Side (ES) of a
Ring Arbiter in HA mode.
Description of States
The state machine of FIG. 9 has the following states.
TABLE-US-00009 TABLE I Number State Description 900 PORT DOWN The
port is operationally down or has just been initialized. Entered
from any state. 910 BLOCKING The node is sending HELLOs, but not
forwarding data traffic. 920 SLAVE FORWARDING The port on a Slave
Node is forwarding packets 930 MASTER The port on a Master Node is
FORWARDING forwarding packets
Description of State Transitions
The table below describes the transitions of the state machine
shown in FIG. 9.
TABLE-US-00010 TABLE J Number Event Action 1910 port operationally
down OR init block user traffic, cancel age timer 1915 port
operationally up start age timer 1920 (age timer expires AND
isMaster) forward user traffic OR RxHello < Node 1925 (age timer
expires AND isSlave) forward user traffic 1930 RxHello < Node
continue forwarding user traffic 1935 RxHello > Node AND
continue forwarding user PartnerHelloTimeout traffic 1940 RxHello
> Node AND NOT block user traffic, start PartnerHelloTimeout age
timer 1945 RxHello > Node AND NOT block user traffic, start
PartnerHelloTimeout age timer
As shown in the sequence of events reported in the table below, the
RS ports of the Arbiters are always forwarding, unless the ports
are OPER DOWN. The point of interest is the state of the ES port of
the Slave Arbiter. In essence, the ES Slave port is FORWARDING if
there is a HELLO timeout on either the RS or ES.
TABLE-US-00011 TABLE K Port Status (before trigger) 500 500 530 530
TIME RS ES RS ES Trigger Reaction 1 800 900 800 900 500 initialized
500 RS goes Blocking, 500 ES Goes to Blocking 2 810 910 800 900 530
initialized 530 RS goes Blocking, 530 ES Goes to Blocking 3 810 910
810 910 500 receives HELLO from 530 500 RS state change 1820 to and
RxHello < node Master Forwarding 500 ES state change 1920 to
Master Forwarding 4 810 910 830 930 530 received HELLO from 500 530
RS state change 1825 to Slave and RxHello > node Forwarding 530
ES does not leave Blocking unless RS or ES has HELLO timeout 5 820
910 830 930 Continues operation with the virtual break in the HA
ring at the ES of the Slave (RA 500). 6 820 910 830 930 Break in
link 1520 (ring side) 7 820 910 830 930 RxHellos stop coming on RS
500 RS no change 530 ES state change 1925 to Slave Forwarding 8 820
920 830 930 All four ports forward traffic while there is a
physical break 9 820 920 830 930 Break fixed 10 820 920 830 930
RxHello received at 530 RS and > node 530 ES state change 1945
to blocking 11 820 920 830 930 Continues operation with virtual
break. 12 820 910 830 930 Link break ES 13 820 910 830 930 HELLOs
stop on ES side of both 530 ES notes that its age timer RA units
expires and it isSlave and has state change 1925 to Slave
forwarding 14 820 920 830 930 All four ports forward traffic while
there is a physical break 15 820 920 830 930 Break fixed 16 820 920
830 930 530 ES receives RxHellos > node 530 ES moves along state
and not transition 1945 to Blocking PartnerHelloTimeout 17 820 910
830 930 Until next break, port down, or switch in node numbers
sufficient to change Master/Slave relationship.
ALTERNATIVE EMBODIMENTS
Unidirectional OPER DOWN Break on Full or High-Availability
Rings
The control system described above assumes that a break in a
network ring will be a bidirectional break as it connects both the
clockwise and counterclockwise virtual breaks upon failure to
receive a HELLO. This bidirectional response could cause a loop in
the event of a unidirectional failure.
FIG. 10, adds additional detail to the drawing shown in FIG. 2.
More specifically, the links are shown in their unidirectional
components rather than as bidirectional links.
For example, when the network ring is fully operational, and the
Slave Arbiter has imposed a virtual break on the west ring port,
Master Arbiter 1000 can receive HELLOs from Slave Arbiter 1025 via
link 11037, Relay 1020, link 11027, Relay 1010, and link 11012.
Likewise, Slave Arbiter 1025 can receive HELLOs from Master Arbiter
1000 via link 11010, Relay 1010, link 11025, Relay 1020, and link
11035.
If link 11027 was cut but link 11025 was left in service, then the
west port on Master Arbiter 1000 would soon stop receiving HELLOs
from Slave Arbiter 1025, while west port of Slave Arbiter 1025
continued to receive HELLOs from Master Arbiter 1000. In the
previously described embodiment, this unidirectional cut at link
11027 would not trigger the Slave Arbiter 1025 to unblock as it
continues to receive HELLOs from Master Arbiter 1000 as these
HELLOs can still travel across intact link 11025. Thus Ring Relay
1020 as well as connected subscriber ports 1045 would be unable to
send data to any node on the ring as Slave Arbiter 1025 is still
blocking data. (Note: as link 11025 is intact, Ring Relay 1020 and
subscriber ports 1045 can still receive data packets)
One alternative embodiment is to react to a port going to an OPER
DOWN state by stopping the transmission of HELLOs and all data from
that port in the opposite direction, effectively creating an
imposed unidirectional break in the other direction. Hence when
ring relay 1010 observes an OPER DOWN associated with link 11027,
Ring Relay 1010 would stop sending HELLOs and all data on link
11025. After Slave Arbiter 1025 fails to receive HELLOs in an
allotted time, the Slave Arbiter 1025 would remove the virtual
break on its west side to allow data traffic from link 11035 to
proceed towards link 11030 or the user ports 1060 and to allow
traffic from link 11032 or user ports 1060 to proceed onto link
11037, thus putting Ring Relay 1020 back into data communication
with the other ring nodes.
Another Solution to Unidirectional Breaks on Full or
High-Availability Rings
In some cases, the OPER state of a ring port may not be DOWN, event
though a unidirectional failure is present. One example of this
type of situation occurs when a leased TDM link (such as a T1, DS3,
or OC-n) is interrupted by the service provider such that there is
no physical layer fault. Likewise, a Gigabit Ethernet link with
auto-negotiation disabled may not report link failure in the case
of a unidirectional break. In either case, no data can traverse the
link in one direction.
A more comprehensive solution to the risk of a unidirectional break
is one that reacts to either an OPER DOWN or an extended break in
the receipt of HELLO messages. In this embodiment a Ring Relay
would respond to either an OPER DOWN on a receiver or a gap of more
than a set time period without receiving a HELLO on a particular
receiver to declare a link failure. For example, a gap of 30
milliseconds (three ticks of a ten millisecond clock) is used in
one embodiment as the indication of too long a period without a
HELLO on a receiver link.
After detecting either type of failure, a Ring Relay does the
following:
Creates a bidirectional break by blocking data on the outgoing link
on the same port as the receiver link deemed to have failed. (Note
that the Ring Relay does not know whether the break discerned on
the receiver link is a bidirectional break or a unidirectional
break but handles all breaks as a unidirectional break. Thus the
Ring Relay may impose a redundant break on a broken transmitter
side link but this is harmless.)
Creates a break at the receive link that is deemed to have failed.
This seems redundant but imposing a break on a broken link allows
the imposed break to remain for a period after the real break is
removed so that the Slave Arbiter has a chance to re-impose a
virtual break before the imposed break is removed at the Ring
Relay.
Sends flagged HELLO protocol packets on the outgoing link on the
same port as the receiver link deemed to have failed. These flagged
HELLO protocol packets serve two purposes. They serve as HELLO
protocol packets so that other nodes do not experience (or
maintain) the state of a HELLO timeout. These flagged HELLO
protocol packets are distinguishable from normal Arbiter generated
protocol packets and thus convey to the Slave Arbiter that a
bidirectional break is being imposed and thus the Slave Arbiter
needs to remove the virtual break.
Blocks HELLO messages from the transmit link on the partner port
relative to the port with the detected failure on the receive link.
Sends flagged HELLO messages on the transmit link on the partner
port. Note that as Ring Relays are provided with generic control
logic which does not need to be configured to indicate the
direction of the Slave Arbiter, the Ring Relay operates without
needing to know which direction is the Slave Arbiter and which
direction is the Master Arbiter. Thus, upon detecting a failure,
the Ring Relay sends flagged HELLO messages in both directions to
ensure that the Slave Arbiter receives the indication that the
Slave Arbiter needs to remove the virtual break.
In the example using FIG. 10, when Ring Relay node 1010 experiences
a HELLO timeout with no OPER DOWN on receiver link 11027, Ring
Relay node 1010 will block all user traffic onto link 11025 and
from link 11027 to thus make a what might be a unidirectional break
on 11027 into a bidirectional break. The Ring Relay 1010 will send
out flagged HELLO messages on both link 11012l and onto link 11025
with the "regenerate" flag set (the normal HELLO messages from
Master Arbiter 1000 will be blocked at 1010 so only the flagged
HELLO messages are placed onto transmit link 11025). The Slave
Arbiter 1025, will see the HELLO messages with the "regenerate"
flag set, and unblock whichever side is currently blocked (east
side port or west side port).
Since each node is regenerating upstream and downstream HELLO
messages in the event of an upstream HELLO timeout, a node may be
assured that it is directly attached to the segment with the
unidirectional failure into its receiver if it experiences a
persistent HELLO timeout. The node directly attached to the segment
that is experiencing such a receiver failure is responsible for
turning the unidirectional break into a bidirectional break and
maintaining that break until the failure is cleared and the Slave
Arbiter has been given time to re-impose the virtual break.
Note that each Ring Relay downstream from the node adjacent to the
break may briefly experience a HELLO timeout before it receives the
flagged HELLOs. During the brief period the downstream nodes may
also impose a break. However, each node will remove the break once
it sees the flagged HELLOs. Also, note that HELLO timeout period
(30 ms) is longer than the maximum time to remove the break (10
ms), ensuring that excess virtual breaks do not propagate upstream
during this brief period. To illustrate this point, if a
unidirectional break that did not generate an OPER DOWN was
incurred on link 11010, then after 30 milliseconds, both Ring Relay
1010 and Ring Relay 1020 will note the absence of HELLO messages
and take corrective action to impose a break and send flagged HELLO
messages. However, when Ring Relay closest to the real failure (in
this case Ring Relay 1010) sends out flagged HELLO protocol packets
on link 11025, these HELLO messages will be received by Ring Relay
1020 which will stop imposing a break on links 11027 and 11025.
When a Ring Relay comes back after an OPER DOWN or a HELLO timeout
failure, the Ring Relay resumes transmissions of normal HELLO
messages first then there is an imposed delay (fd as discussed
above) before removing the virtual breaks for data on both the
receiver link deemed to have failed and the transmit link on that
same port to avoid having a temporary loop before the Slave Arbiter
has a chance to impose a virtual break.
The Master Arbiter must be modified to incorporate rules used by
the Ring Relays as the Master Arbiter acts much like a Ring Relay
with respect to reacting to a lack of HELLOs or an OPER DOWN on a
receiver link for the Master Arbiter. More specifically, the Master
Arbiter must be adapted to suppress transmission of user data and
set the regenerate bit on transmitted HELLOs on a port that is
experiencing HELLO timeouts or an OPER DOWN. In this case the
Master Arbiter may surmise that the unidirectional break is present
on its directly attached segment. If flagged or original HELLOs are
received, the Master Arbiter transmits HELLOs with no regenerate
flag set and passes user data as usual after a period of delay
sufficient to allow the Slave Arbiter time to reinsert the virtual
break and thus avoid a temporary loop. A Master Arbiter modified to
act like a Ring Relay would respond to a failure by sending flagged
HELLOs out both ports.
One minor difference between a Master Arbiter and a Ring Relay is
that the Master Arbiter knows that the Slave Arbiter is both on the
east side and on the west side of the Master Arbiter. In contrast
as a Ring Relay does not ordinarily send HELLO messages through a
Master Arbiter. the Slave Arbiter is only to one side of the Ring
Relay, not both. In a preferred embodiment the Master Arbiter is
configured to transmit flagged HELLOs just on the port with the
receive link deemed to have failed. The partner port continues to
send normal HELLOs
The following example uses the network loop shown in FIG. 10 but
assumes that all links are initially working properly. When link
11012 fails through either an OPER DOWN or a problem that causes
Master Arbiter 1000 to note that the west port is not receiving
HELLOs, a Master Arbiter configured to send flagged HELLOs on link
11010 but continues to send normal HELLO messages on its partner
port may not actually provide notice to Slave Arbiter 1025 of a
problem as there may have been a bidirectional break rendering link
11010 inoperative. More specifically. when a bidirectional failure
is experienced at links 11012 and 11010, then the Master Arbiter
would impose a virtual break on the west port (as it does not know
whether the break is a unidirectional break link 11012l only or a
bidirectional break as generally it is not possible to perceive a
break on a transmit link) but continue to send flagged HELLOs on
transmit port 11010. As these flagged HELLOs cannot reach Ring
Relay node 1010 across broken link 11010, Ring Relay node 1010 will
also note the gap in HELLOs and will take steps to impose a virtual
break on links 11012 and 11010. When Ring Relay 1010 responds to
the gap in HELLOs or OPER DOWN a portion of the response will
include the sending of flagged HELLO protocol packets on link
11025. These flagged HELLO protocol packets from Ring Relay 1010
ultimately reach the Slave Arbiter and trigger the removal of the
virtual break at a port connected to Slave Arbiter 1025.
The Slave Arbiter has two tasks. The first task is like other
nodes, it converts sensed link failures (either OPER DOWN or HELLO
Timeout) as a potential unidirectional break and imposes a break on
the transmit link on the port with the perceived failure of its
receive link. The Slave Arbiter has the second task of removing the
imposed virtual break that is used to avoid looping after another
bidirectional break exists in the network loop.
The methods of implementing these two tasks could be done in
several different ways depending on a design choice to use common
instructions for both Master Arbiter and Slave Arbiter in their
response to what may be a unidirectional failure. Using common
instructions for this aspect of operation is potentially useful as
in the preferred embodiment an Arbiter contains the instruction set
necessary to act as either a Master Arbiter or a Slave Arbiter.
Assuming that the design choice is made to use common instructions
with the Master Arbiter, the sequence of events for a break
detected by a Slave Arbiter is illustrated by the following
examples using the components shown in FIG. 10. In each example,
the example starts with all links functioning (including link
11027) and with a virtual break imposed on the east port of Slave
Arbiter 1025 (links 11030 and 11032 passing HELLOs but not user
data).
First Case--Break on Incoming Link on Port Away from Virtual
Break
Link 11035 experiences a failure that does not lead to an OPER DOWN
but does lead to a HELLO timeout noted at Slave Arbiter 1025. The
Slave Arbiter acts like any other node to impose a bidirectional
break at the failed port. The Slave Arbiter imposes a break on
links 11035 and 11037. If the Slave Arbiter is using common
instructions with the Master Arbiter mode, then the Slave Arbiter
would send flagged HELLOs instead of HELLOs on link 11037 despite
the fact that this flagged HELLO is only of significance to the
Slave Arbiter.
The recovery in this first case would start when the problem with
link 11035 is corrected and normal HELLOs from Master Arbiter 1000
are again received at the Slave Arbiter on link 11035. A Slave
Arbiter using common instructions with a Master Arbiter would stop
sending flagged HELLOs and would resume sending normal HELLOs. The
Slave Arbiter would re-impose the virtual break so the virtual
break is in place after the imposed bidirectional break is removed.
After a delay of fd, the Slave Arbiter would remove the imposed
bidirectional break.
Second Case--Bidirectional Break between Ring Nodes
Link 11025 experiences a failure that does not lead to an OPER DOWN
but does lead to a HELLO timeout noted at Ring Relay node 1020. The
problem is actually bidirectional so that Ring Relay 1010 also
experiences a HELLO timeout on link 11027. As described above, Ring
Relay node 1020 would create a bidirectional break on the port
experiencing the receiver failure so that no data passes through
that port but sends out flagged HELLOs onto link 11027 and listens
for HELLOs on link 11025. Ring Relay 1020 sends out flagged HELLOs
on both link 1027 and link 11035.
As Ring Relay 1010 is also experiencing a HELLO timeout on link
11027, this Ring Relay imposes a bidirectional break on links 11027
and 11025 and sends out flagged HELLOs on links 11025 and link
11012.
The Slave Arbiter would start the process described in the First
Case of imposing a bidirectional break on links 11035 and 11037 in
addition to the existing virtual break on the east port since the
Slave Arbiter would have no way of knowing if the problem was in
the link adjacent to the Slave Arbiter or upstream from there. The
flagged HELLOs sent on link 11035 are received at Slave Arbiter
1025 which then removes the virtual break on links 11030 and 11032
as the receipt of a flagged HELLO indicates that a bidirectional
break is being imposed by another node on the network ring. After a
delay of fd, the Slave Arbiter would remove the imposed
bidirectional break on links 11037 and 11035. After this sequence
of events the network ring would have an imposed bidirectional
break on links 11025 and 11027 at Ring Relay 1020 and again at Ring
Relay 1010.
When the problem on links 11025 and 11027 are resolved, Ring Relay
nodes 1010 and 1020 will stop sending flagged HELLO and then after
a delay remove the imposed bidirectional breaks on links 11027 and
11025. Slave Arbiter 1025 upon receipt of a normal HELLO from
Master Arbiter 1000 will impose a virtual break on the port in
Master Forwarding mode which in this situation should be the east
port (links 11030 and 11032).
Third Case--Break on Outgoing Link on Port away from Virtual
Break.
Link 11037 experiences a failure that does not lead to an OPER DOWN
but does lead to a HELLO timeout noted at Ring Relay 1020. As noted
above, Ring Relay 1020 will impose a bidirectional break on links
11037 and 11035 and start sending flagged HELLOs on links 11035 and
11027. (As the temporary imposition of additional bidirectional
breaks by downstream nodes such as Ring Relay 1010 and Master
Arbiter 1000 has already been explained above, it will be omitted
in subsequent discussions). The Slave Arbiter will remove the
virtual break at links 11030 and 11032 after receiving a flagged
HELLO on link 11035. When the problem with link 11037 is resolved,
the Ring Relay 1020 will stop sending flagged HELLOs and will allow
normal HELLOs originating from Master Arbiter 1000 to reach the
Slave Arbiter 1025 on link 11035. The response to the receipt of
normal HELLOs instead of flagged HELLOs by the Slave Arbiter 1025
is to re-impose the virtual break on links 11030 and 11032.
Fourth Case--Break on Incoming Link at Virtual Break
Link 11030 experiences a failure that does not lead to an OPER DOWN
but does lead to HELLO timeout at Slave Arbiter 1025. Although link
11030 has an imposed virtual break for data traffic. HELLOs are
received at Slave Arbiter 1025 and the lack of HELLOs triggers a
response. Functionally, a bidirectional break is imposed on links
11030 and 11032 (including the transmission of flagged HELLOs if
the Slave Arbiter is configured to use the same code as the Master
Arbiter), then after a short delay, the virtual break is removed
from links 11030 and 11032. One of skill in the art can recognize
that the logic could be modified to simply leave the virtual break
in place but note that the reason for the break has changed from
the Slave Arbiter needing to impose a virtual break to the need to
convert a potential unidirectional break into a bidirectional
break.
After the problem on link 11032 is corrected, the Slave Arbiter
receives normal HELLO messages from Master Arbiter 1000 and
reverses the process by first imposing a virtual break on links
11030 and 11032 and then removing the imposed bidirectional break
on links 11030 and 11032. Again, one of ordinary skill in the art
can see that this can be done by changing the reason for the break
without adding and removing breaks.
Fifth Case--Break on Outgoing Link at Virtual Break
Link 11030 experiences a failure that does not lead to an OPER DOWN
but does lead to a HELLO timeout at Ring Relay 1015 (and as
discussed above, temporarily at Ring Relay 1005 and Master Arbiter
1000). To distinguish this example from the last case, assume that
the problem is unidirectional such that link 11032 is operating
properly and Slave Arbiter 1025 is receiving HELLOs on link 11032.
Ring Relay 1015 imposes a bidirectional break on links 11030 and
11032 and starts to send flagged HELLOs on links 11032 and 11020.
This bidirectional break imposed by Ring Relay 1015 is redundant to
the virtual break already imposed by the Slave Arbiter as both
block data but not HELLOs. The receipt of a flagged HELLO at link
11032 of Slave Arbiter 1025 indicates that another node has imposed
a bidirectional break so the virtual break is removed from 11030
and 11032 by the Slave Arbiter. While the removal of a virtual
break from links 11030 and 11032 has no practical effect, one can
appreciate that the Slave Arbiter does not know whether the flagged
HELLOs are coming from an immediately adjacent node (Ring Relay
1015) or a remote node (such as Ring Relay 1005 or Master Arbiter
1000). If the flagged HELLOs were coming from a remote node, then
removal of the virtual break would make the adjacent node (in this
case Ring Relay 1015) connected for data communication with the
rest of the ring.
When the problem with link 11030 is corrected, Ring Relay 1015 will
receive normal HELLO messages from Slave Arbiter 1025. The receipt
of normal HELLO messages will cause the Ring Relay to stop sending
flagged HELLO messages. The receipt of a normal HELLO message at
link 11032 of Slave Arbiter 1025 will cause the Slave Arbiter to
re-impose a virtual break on links 11030 and 11032. After a short
delay, Ring Relay 1015 removes the bidirectional break from links
11030 and 11032.
Separation of HELLO Message and Imposed Break Message
One of ordinary skill in the art can see that a flagged HELLO
protocol packet as discussed above actually serves two purposes.
First purpose is to provide a HELLO message to downstream nodes so
that the HELLO timers do not see a gap in HELLOs. The second
purpose is to convey a message to the Slave Arbiter that another
node has imposed a bidirectional break so the Slave Arbiter can
remove the virtual break. A modification to the disclosed
embodiment would be to pass (and create if necessary) HELLO
messages whenever the preferred embodiment calls for a flagged
HELLO and to augment that with another control signal message that
indicates that a device has imposed a bidirectional break. The
Slave Arbiter would react to the control signal message to remove
the virtual break. One of ordinary skill in the art will recognize
that a system using a separate control signal in lieu of the field
in the flagged HELLOs could communicate to the Slaver Arbiter the
need to re-impose the virtual break either by ceasing to send a
control signal to remove the virtual break or by sending a control
signal to re-impose the virtual break.
While discussing the flagged HELLO message it is worth noting that
flagged HELLO protocol packets are sometimes called regenerated
HELLO protocol packets or regenerated HELLO messages as ring relays
"regenerate" HELLOs when they do not receive them. However as not
all of these flagged HELLO messages are actually "regenerated,"
flagged is a better descriptor that regenerated. The term "flagged"
should not be read as limiting the way of distinguishing one type
of HELLO from another to setting a flag value, although that is a
suitable solution. Any readily discernible difference between the
normal HELLOs and the flagged HELLOs such that the Slave Arbiter
can discern the information conveyed by the use of the flagged
HELLO is sufficient.
Unidirectional Transmission of Flagged HELLOs
Another variation on the preferred embodiment differs at the Ring
Relay in that the Ring Relay would send flagged HELLOs out the
partner port of the port with the failure on the incoming link but
the outgoing link on the port with the failure on the incoming link
would continue to get whatever HELLOs were received by the Ring
Relay. For example, using FIG. 10 with all links operating
including 11027, if Ring Relay 1020 stopped receiving HELLOs on
incoming link 11037, then Ring Relay 1020 would send flagged HELLOs
out link 11027 but would not replace the HELLOs received on link
11025 with flagged HELLOs for transmission on link 11035. This
reduces the actions necessary at Ring Relay 1020.
In order for this variation to reliably provide the flagged HELLOs
to the Slave Arbiter 1025, the Master Arbiter 1000 would need to be
modified to respond to the receipt of a flagged HELLO from an
incoming link by reflecting back a flagged HELLO on the
corresponding outgoing link on that same port. In this case a HELLO
timeout on link 11037 would lead to the creation of a flagged HELLO
at Ring Relay 1020 which would be passed over link 11027 to Ring
Relay 1010 and then over link 11012l to Master Arbiter 1000. Master
Arbiter 1000 would respond by sending a flagged HELLO out link
11010 to Ring Relay 1010 then over link 11025 to Ring Relay 1020
and finally over link 11035 to Slave Arbiter 1025. In the event
that the problem at 11037 was part of a bidirectional problem
involving link 11037, then Slave Arbiter 1025 would not receive the
flagged HELLO from Master Arbiter 1000 but would respond to a HELLO
timeout and impose a bidirectional break on its own.
Rings without a Master Arbiter
In yet another embodiment, Master Arbiter 1000 shown in FIG. 10
(with link 11027 assumed to be working in this example) could be
removed from the ring so that Ring Relay 1010 is connected directly
to Ring Relay 1005. In this embodiment, the Slave Arbiter 1025
would be the only source of normal HELLOs. Each ring relay could
send out flagged HELLOs in just the direction with the incoming
link having a problem. For example, if incoming link 11025 for Ring
Relay 1020 experienced a HELLO timeout as part of a unidirectional
failure, then Ring Relay 1020 could send flagged HELLOs (or some
other control signal) out 11027 and the flagged HELLO would reach
the Slave Arbiter. The Slave Arbiter would react to an OPER DOWN or
a lack of HELLOs as described above. In this embodiment, there is
no need for the Slave Arbiter to switch from sending HELLOs to
flagged HELLOs (those with the control signal noting that the Slave
Arbiter is imposing a bidirectional break on a port) as only the
Slave Arbiter reacts to a difference between a HELLO and a flagged
HELLO.
Dual Homing Using a Single Node Ring
FIG. 11 shows an application of a particular embodiment of the
present invention that is referred to as "dual homing". Dual homing
allows a Slave Arbiter Node 1130 to provide protected access for
User Ports 1140 to network via Ring Access Equipment nodes 1110 and
1120 using redundant links 1160 and 1170.
In this alternative embodiment, the Slave Arbiter node 1130 would
see its own HELLOs. As described in Table C, one side of 1130 (for
example the west side of the Slave Arbiter connected to link 1160)
would go to the SLAVE FORWARDING state and one side (for example,
the east side of the Slave Arbiter connected to link 1170) would go
to the BLOCKED state.
Now, in response to a fault on the Ring Access Equipment 1110 or
the link 1160, the east side of the Slave Arbiter 1130 would
unblock, and the User Ports 1140 would continue to have access to
the network. The network access for User Ports 1140 is therefore
protected against faults in either the access links (1160 and 1170)
as well as in the Ring Access Equipment nodes (1110 and 1120).
Unidirectional Break for Dual Homing Using a Single Node Ring
FIG. 12 is another view of a Dual Homing application. Dual homing
allows a Slave Arbiter Node 1230 to provide protected access for
User Ports 1231, 1232, and 1233 to network via Ring Access
Equipment nodes 1210 and 1220 using redundant links. In order to
discuss a unidirectional break, the redundant links are shown in
using the various unidirectional components: 1260, 1261, 1270, and
1271.
Through the use of FIG. 12, it is possible to discuss an embodiment
of a dual homing ring experiencing a unidirectional break. In this
embodiment the following rules are followed; No OPER DOWN and
receiving HELLOs on both ports: Impose a virtual break on one
bidirectional port as discussed in connection with FIG. 11 to block
data packets but not control messages such as HELLOs. Experiencing
a HELLO timeout on both ports or a HELLO timeout on one port and an
OPER DOWN on the other port (indication a problem in Ring Access
Equipment 1210 or 1220 or a bidirectional break: Remove the virtual
break from the port with the imposed virtual break so that user
ports can have access to the network through either access point
(if either is working)). Experiencing an OPER DOWN or a HELLO
timeout on one port but not the other (indicating a unidirectional
break): On the port with a failed receiver (either OPER DOWN or no
HELLOs) impose a bidirectional break so that the virtual break can
be removed without causing a unidirectional loop. After the
bidirectional break is imposed, then remove the virtual break.
After the problem with the receiver is corrected, then impose the
virtual break before removing the bidirectional break.
For example, consider the case where 1230 is initially imposing a
virtual break on its east side (links 1270 and 1271). Upon a HELLO
timeout on link 1261, Slave Arbiter 1260 it would impose a
bidirectional break on links 1261 and 1260 and remove the virtual
break on links 1270 and 1271. After the problem with link 1261 was
corrected, the virtual break would be imposed on links 1270 and
1271 and the imposed bidirectional break on links 1260 and 1261
would be removed a short time later. One of ordinary skill in the
art will recognize that the imposition of a virtual break could be
done on the opposite port from the port which most recently had the
virtual break without deviating from the teachings of the present
application.
If the lack of HELLOs was on the side with the imposed virtual
break then the Slave Arbiter 1230 would impose a bidirectional
break on links 1270 and 1271 before removing the pre-existing
virtual break on links 1270 and 1271. Upon receipt of HELLOs on
link 1270, the process would be reversed with the imposition of a
virtual break on links 1270 and 1271 before removing the
bidirectional break on links 1270 and 1271. One of ordinary skill
in the art could modify to the control rules to simply maintain the
imposed break if called for by the rules for virtual breaks or if
called for by the rules for bidirectional breaks rather than
imposing and removing redundant breaks.
There is no need for flagged HELLOs in this situation as the only
device that makes use of the distinction between a normal HELLO and
a flagged HELLO is the slave arbiter that is sending the
HELLOs.
Use of a Control Message in Place of the fd Timer
The preferred embodiments disclose using a timing delay to ensure
that a port progressing from OPER DOWN to operational delays
removal of the bidirectional break long enough for the Slave
Arbiter to impose a virtual break. Likewise a port that had
experienced a HELLO timeout delays removing the bidirectional break
long enough for the Slave Arbiter to impose a virtual break. One of
skill in the art will recognize that the use of the timer could be
replaced by a control signal sent by the Slave Arbiter after it has
successfully imposed the new break before removing the previously
existing break. Analogous to the regenerate flag in the flagged
HELLO message, this control signal could be a flag set in the HELLO
protocol packet sent by the Slave Arbiter.
One of skill in the art will recognize that alternative embodiments
set forth above are not universally mutually exclusive and that in
some cases alternative embodiments can be created that implement
two or more of the variations described above.
Those skilled in the art will recognize that the methods and
apparatus of the present invention have many applications and that
the present invention is not limited to the specific examples given
to promote understanding of the present invention. Moreover, the
scope of the present invention covers the range of variations,
modifications, and substitutes for the system components described
herein, as would be known to those of skill in the art.
The legal limitations of the scope of the claimed invention are set
forth in the claims that follow and extend to cover their legal
equivalents. Those unfamiliar with the legal tests for equivalency
should consult a person registered to practice before the patent
authority which granted this patent such as the United States
Patent and Trademark Office or its counterpart.
ACRONYMS
ES Extension Side
FR Full Ring
HA High Availability
IP Internet Protocol
MAC Media Access Control
MPLS Multiprotocol Label Switching
PDU Packet Data Unit
PSR Protected Switching Ring
RA Ring Arbiter
RPR Resilient Packet Ring
RR Ring Relay
RS Ring Side
TCP Transmission Control Protocol
UDP User Datagram Protocol
* * * * *
References